See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/384322522

Pressure-Driven Batch Distillation Optimal Control for Acetone- Methanol

Separation

Article in Fine Chemical Engineering · September 2024

DOI: 10.37256/fce.5220244485

CITATIONS READS

0 2

All content following this page was uploaded by Marija Stojkovic on 30 September 2024.

The user has requested enhancement of the downloaded file.

Fine Chemical Engineering
http://ojs.wiserpub.com/index.php/FCE/

Research Article

Pressure-Driven Batch Distillation Optimal Control for Acetone-

Methanol Separation

Marija Stojkovic

10 Copenhagen Street, 53100 Mayenne, Loire County, France

E-mail: [email protected]

Received: 20 February 2024; Revised: 9 September 2024; Accepted: 19 September 2024

Abstract: Over the last fifteen years, immense progress has been made in the research of pressure-swing batch
distillation. The challenge lies in the fact that certain pressure-sensitive azeotropic mixtures cannot be separated in
a regular open batch mode with an acceptable outcome. Throughout most of this text, findings contradict previously
established facts. The separation of acetone-methanol by pressure-swing batch distillation in a mixed double system,
consisting of a regular and inverted column, is the process under investigation. In this work, a comprehensive global
solution to the optimal control problem is derived in the form of a sequential synthesis of controlled trajectories.
A thorough analysis of the control variables, with and without parametric sensitivity of manipulated variables, on
the optimal control pattern with variable reflux is presented. Most authors did not achieve significant results by
simultaneously optimizing reflux and liquid division ratios or by using a “variable pressure gap.” This study challenges
previous findings, emphasizing the importance of simultaneous optimization of all three factors mentioned. During
this study, the optimal reflux strategy through cyclic operation was extended to optimize energy expenditure over a
fixed time horizon. Moreover, the proposed scheme offers a high level of operational flexibility, allowing units to be
operated independently or as a system. Additionally, the facility of connecting additional devices is highlighted by the
significant difference in output temperatures and the potential for using evaporated fluids consecutively. Lastly, the
research provides a solid basis for future investigations into internal and external heat integration, potentially leading to
conclusions about expanding the network using well-known heat cycles.

Keywords: pressure-swing, batch distillation, optimal control, direct method, cyclic distillation

Nomenclature
AIB Advanced inverted batch with an additional tank at the top
ANN Artificial Neural Network
BB Big bottoms
BE Batch extractive column
BR Batch rectifier
BS Batch stripper
DCBR Double column batch rectifier

Copyright ©2024 Marija Stojkovic.

DOI: https://doi.org/10.37256/fce.5220244485
This is an open-access article distributed under a CC BY license
(Creative Commons Attribution 4.0 International License)
https://creativecommons.org/licenses/by/4.0/

Fine Chemical Engineering 498 | Marija Stojkovic

DCBS Double column batch stripper
DEG Diethyleneglycol
DHSA Distillation harmony search algorithm
EAPSD Entrainer assisted pressure-swing distillation
GA Genetic algorithm
HHIPSD Hybrid heat-integrated pressure-swing distillation
MEG Mono ehtylene glycol
MBDS Multi Batch Distillation Software
PSOA Particle swarm optimization algorithm
RB Regular batch
SIM Sequential iteration method
SMBO Surrogate model-based optimization
SQP Sequential Quadratic Programming
TAC Total annual cost
TCC Triple column configuration
THF Tetrahydrofuran

1. Introduction
Repke and Klein1 were the first to propose a simple control scheme for a pressure-swing inverted batch double-
column system: product concentration was controlled by the reflux rate flows. Indeed, the low-pressure column showed
a rapid response, with a jump from third to second decimals. Repke et al.2 recognized the volumetric ratio against molar-
decisive and recommended the inverted configuration in the case of the light component if the initial feed is lower than
0.25. Modla and Lang,3 however, compared different configurations to conclude: (1) single/double rectifier are only
suitable for maximum boiling azeotrope, (2) single/double inverted column are only suitable for minimum boiling point
azeotrope, (3) a combination of the rectifier and inverted column is possible to use only if in each step the azeotrope
is continuously withdrawn as a product, (4) middle vessel is not suitable for any. Later, Modla and Lang4 introduced
a “speed of pressure change” to end up with conclusions about the products obtained, respectively: (1) for maximum
boiling azeotrope, double batch rectifier was the most “recovering” configuration, giving 7.5%/10.6% more concerning
the combination of rectifier and stripper, and only 1.5%/6.4% more for a single rectifier, (2) for the minimum boiling
azeotrope, double stripper shown to be the best “recovering” configuration giving 3.3%/29.2% more to the combination
of rectifier and stripper, respectively, and 19.8%/11.9% to a single stripper. But Modla5 established as follows: (1)
minimum boiling azeotrope separation is feasible in all except closed stripper, (2) maximum boiling azeotrope
separation is feasible in closed stripper and open rectifier.
Even though their work involves classical control techniques, according to Modla,5 Li et al.6 and Kopacz et al.,7 it
is worth mentioning the findings. Kopacz and coworkers8 introduced a vapor division ratio which proved to affect the
reflux ratio and consequently total time/energy requirement. Modla5 again highlighted the significance of the influence
of vapor division ratio on energy requirement: double-column rectifier in closed mode resulted in 3.2/1.74 times more
reduction of energy requirement compared to the open mode/single inverted column, respectively, but double stripper
configuration in the open mode required even 1.54 times more energy compared again to the single open stripper. Modla
and Lang9 (DCBS open mode) examined the process with/without heat integration: having pressure gap increased by 9
times, CO2 emission reduction increased by 19.12%/29.05%, respectively. Modla,10 for the charge composition inside/
outside of the pressure-swing operating region, claims the decrease in energy expenditure by (29.85%/49.56%)/37.42%,
for DCBR/TCC against the (BS-DCBR/TCC)/(BS-DCBR-DCBR), respectively. Further, Modla and Lang11 proved that
thermal integration benefits go up to 59.18%/67.13%, for the pressure-swing/extractive processes, respectively. Modla12
studied a system comprised of the reactive rectifier combined with a non-reactive stripper to discover a “sharp uprisal”
in energy demand, for feed plate locations ranging from 35 to 40.
Even with classical control techniques applied, the results of Li et al.13 and Wang et al.14 are noteworthy. It is to
note that, after having combined a system of double column stripper batch with a continuous rectifier, Li et al.13 proved a

Volume 5 Issue 2 |2024| 499 Fine Chemical Engineering

minimum pressure gap of 4.4 atm and succeed in impure alcohol products to their maximum. On the other side, having
low/high-pressurized column pressure increase by 39.50%/75%, respectively, Wang et al.14 gained 12.11%/19.93% in
TAC reduction. Compared with the improvement in terms of greenhouse gas emissions, high-pressurized columns saved
almost 10% more. Most recently, for a minimum-boiling azeotrope of THF-water, Luyben,15 using a double continuous
mode rectifier with vapor recompression, delimited LP/HP working pressure conditions to a vacuum of (0.6 atm)/above
technical (11.3 bar): This outperformed “traditionally researched scheme” (Luyben16), by 19.4% in TAC, whereas, by
40.15% with heat integration.
Having considered three cases according to the content of the light component, Li et al.6 concluded that: (1) if a
low-pressurized column operated at atmospheric pressure and the pressure of a high-pressurized column is increased
by 27.28%, it influences TAC to decrease by 23.87%, (2) also if low-pressurized column pressure increases only
by 0.57%, the same pressure increase as in the first case for the high-pressurized column here invokes reduction of
TAC by 34.19%. For a particular case examined, it seems that if the pressure gap is reduced by less than 1%, TAC
is, however, reduced by more than 10%. Even with classical control techniques applied, the results of Li et al.13 and
Wang et al.,14 are noteworthy. It is to note that, after having combined a system of double column stripper batch with a
continuous rectifier, Li et al.13 proved a minimum pressure gap of 4.4. atm and succeed in impure alcohol products to
their maximum. On the other side, having low/high-pressurized column pressure increase by 39.50%/75%, respectively,
Wang et al.14 gained 12.11%/19.93% in TAC reduction. Compared with the improvement in terms of greenhouse gas
emissions, where, high-pressurized columns saved almost 10% more. Most recently, for a minimum-boiling azeotrope
of THF-water, Luyben,15 using a double continuous mode rectifier with vapor recompression, delimited LP/HP working
pressure conditions to a vacuum of (0.6 atm)/above technical (11.3 bar): This outperformed “traditionally researched
scheme” (Luyben16) by 19.4% in TAC, whereas, 40.15% with heat integration. Afterward, Luyben17 in particular proved
benefits from vacuum distilling: For a specified range of pressure within the technical limits, for an increase of [2-2.5]
folds, the reduction in capital cost varies from 20.64%-21.22%. Furthermore, for a ternary mixture of MEG/DEG and
water, the author concluded for derived increments on investment, going from higher to lower vacuum pressure, to rise
if moving toward the higher decimal value.
Most recently, Kale et al.18 studied and optimized the batch distillation system of acetone and n-heptane, and found
the optimal parameters to be a reflux ratio of 2.2, the total number of plates to be 19, in the total time interval fixed for 1 h,
to purify more than 99% of acetone from the azeotropic mixture. Ding et al.,19 in their first case, separated acetone from
methanol in a double-column system, to compare different algorithms in terms of the optimal value for TAC: Distillation
harmony search algorithm (DHSA) gave, 0.5%/0.9%/0.6% lower objective than that of sequential iteration method
(SIM)/genetic algorithm (GA)/particle swarm optimization algorithm (PSOA), respectively. Hegely et al. 20 separated
methanol from a 5-component mixture to discover that the influence of the price of incineration on the objective
function, i.e., profit, is significant, because the profit decreases by 78.42% when the price of incineration increases
only by 17.72%. Qin et al.21 used entrainer-assisted pressure-swing batch distillation for separating pressure-insensitive
binary azeotrope: The authors showed that “the order” of the different pressure set-up sequences is also important as
they proved that heat-integrated high-low pressure sequence has lower TAC and CO2 emissions, by 7.04%/4.27%,
respectively, compared to low-high pressure sequence. Miao et al.22 introduced a new hybrid heat-integrated pressure-
swing distillation process with a heat pump to separate both maximum- and minimum-boiling azeotropes, and,
compared with the conventional fully heat-integrated pressure-swing distillation process, to conclude that the first one
gives from 28.8% to 51.9% reductions in TAC, depending on the azeotropic system observed.
In this study, the optimal control for the operation of pressure-swing batch distillation for acetone production in
an open-mode double-column configuration is researched. The goal of this study is to investigate the optimal control
functioning of the combined scheme, composed of both batch rectifier and batch stripper, in “dwelling” between
pressures of atmosphere and high pressure allowed the industrial practice. Throughout the manuscript, firstly the
balances are introduced for column functioning at elevated pressure, then the optimal control problem is defined, where
some important thermodynamic conditions are introduced as an operational constraint since the latter is key for safety
operation(s). From here, the optimal control reflux policy as a cyclic operation is obtained. Furthermore, to research
more in-depth the influence of different parameters on the optimal control structure, the author extends this study by
(1) investigating the influence of liquid ratio, and, the investigation of the influence of vapor boil-up, (2) investigating
the influence of reboiler volume, (3) investigating the influence of the pressure gap. Least, but not last, within the

Fine Chemical Engineering 500 | Marija Stojkovic

investigation of the influence of vapor boil-up and the investigation of the influence of the pressure gap, the optimal
reboiler duties are obtained as discontinuous functions over the fixed time horizon for each investigation. This is of the
most importance, as it shows that it is possible to achieve the reduced energy requirement.
All in one, for the very first time a cyclic operation as an optimal control reflux policy of long-time researched
azeotropic mixture is developed. Table 1, shows a list of studies on the optimal control for homogeneous systems in
discontinuous and/or semi-continuous mode for pressure-swing distillation.

Table 1. A list of different studies on the optimal control for homogeneous systems in discontinuous and/or semi-continuous mode for pressure-swing
distillation

Authors(s) year column (s) method/control mixture

Repke et al.1 2005 BS gPROMS acetonitrile-water

Repke et al.2 2006 BR, BS gPROMS acetonitrile-water

BS, BR; DCBS, DCBR, Ethanol-toluene
Modla and Lang4 2007 CCDCOLUMN
BS+BR water-ethylenediamine
Modla and Lang9 2010 BS ChemCad Acetone-methanol
DCBS
Modla and Lang11 2011 ChemCad Acetone-methanol
BE
Li et al.13 2019 DCBS, TCC SQP/AspenPlus THF-methanol-water
DCBS-series
Zhao et al.23 2021 DCBS-parallel AspenPlus hexane/ethanol/butanone
BS+DCBS
DCBS with decanter Cyclohexane-acetonitrile-
Zhao et al.24 2021 AspenPlus
*heterogeneous toluene
Qin et al.21 2022 EAPSD GA Water-isopropanol
Water-ethylenediamine,
Miao et al.22 2022 HHIPSD GA Methanol-aceetone
Kale et al.18 2023 BR MBDS Acerone-(n-heptane)
Acetone-methanol, Silicon
SIM, GA,
Ding et al.19 2023 DBCR, TCC tetrachloride-trichlorosilane-
PSOA, DHSA dichlorosilane
Acetone-methanol-
ANN, GA,
Hegely et al.20 2023 BR tetrahydrofuran-water-
SMBO, SQP toluene

2. Materials and methods

2.1 Process dynamics model
The following assumptions are taken: (1) equimolar overflow, (2) total condensation, (3) vapor hold-up dynamics
neglected, (4) pressure drop along the columns neglected, and (5) ideal mixing of the vapor and the liquid phase. The
detailed mathematical model developed by Stojkovic et al,25 which includes mass balances on all the parts of the batch
column (MES), is used.

Volume 5 Issue 2 |2024| 501 Fine Chemical Engineering

 V Condenser:
P1 = 1 atm
Tcond = 63.31℃ L
Condenser:
L P2 = 10 atm
Tcond = 138.39 ℃
D
Tank PL < PH
V

V
V-1 Pots
Pump

Reboiler V
Charge:10I
0.95 acetone
0.05 methanol

Pre-Heater Reboiler
Obligatory streams D
Optional streams Tank Tank

Figure 1. A combined scheme for pressure-swing batch distillation

However, having the second column set at high pressure (P > 3 atm) implies the necessity to take the pressure
effects into account. According to Doherty and Malone,26 for high-pressure distillation, the vapor-liquid equilibrium
relation recommended is:

 ( v − Bii )( P − Pi ) 
Pyi = Pγ i xi exp  i  (1)
 RT 

Equation (1), P, Pi refers to total/partial pressure, respectively, whereas, xi, yi refers to the molar concentrations of
the respected component in the liquid/vapor phase, respectively. Furthermore, γi, υi refers to the coefficient of activity/
volume of the pure liquids/second virial coefficients, respectively. The coefficients of activity are described by the NRTL
model, subsequently, the parameters are tabulated in Table 2. In Table 3 and Table 4, however, Antoine’s parameters are
tabulated, for respective components over different intervals of temperature.27 A second virial coefficient denoted as Bii
is calculated from Pitzer-Curl equations for non-polar molecules.28
Overpressuring effects should be avoided at first because if the control valve on a feed line falls fully open, it can
result in overpressure in a reboiler, accumulator, or even main trays. Consequently, the left side of equation (1) must be
constrained to the value of the working pressure during the operation.

Table 2. Thermodynamics data

Binary interaction parameters for NRTL model

Component i Component j Aji [cal/mol] Aji [cal/mol] α
Acetone Methanol 78.317 140.046 0.47
Antoine parameters from DECHEMA
Acetone
n° A B C Tmin, [C°] Tmax, [°C]
1. 7.1327 1,219.97 230.65 -64 70
2. 7.6313 1,566.69 274.42 57 205
Methanol
n° A B C Tmin, [°C] Tmax, [C°]
1. 8.0810 1,582.27 239.7 15 100
2. 7.9701 1,521.23 234 65 214

Fine Chemical Engineering 502 | Marija Stojkovic

 The pressure-swing process scheme presented in Figure 1, assumes “one-pass” through the column at the time [2],
in other word, it is assumed that columns work consecutively (“two-step” process). In step (1) LP run (rectifier), and,
in step (2), HP run (stripper), whereas the maximum achievable distillate concentration at lower pressure P1 = 1 atm is
78% of acetone, but, at P2 = 10 atm only 37% of acetone.

1 1
(a) (b)
0.9 0.9

0.8 0.8

0.7 0.7

0.6 0.6

0.5 0.5

0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Figure 2. Vapor-liquid equilibrium curves for acetone-methanol at different pressures: (a) P1 = 1 atm; (b) P2 = 10 atm

2.2 Optimal control problem

In the same Figure 2, the vapor-liquid equilibrium is presented for a mixture of acetone-methanol at different
pressures: The data obtained from Matlab Version R2017a simulation are plotted against ChemCad 6.1 professional
simulator for the same thermodynamics model, described previously. In Figure 2, on the right/left side plot, respectively,
the vapor-liquid equilibria for pressure set at P1 = 1 atm/P2 = 10 atm, respectively, show that the azeotropic point
coincides with 78%/37%, respectively, containing acetone in the mixture.
Since the set of equations describing the related mathematical model is attributed to the conservation of the
mass property, hereby, the “conservation of the total working pressure” through equation (1) is assured as well: The
summation of partial pressures of components needs to be equal to the working pressure at any moment, whereas to
respect the physical limits imposed by the molar fractions of components in vapor/liquid phase.
Consequently, relations defining physical properties are:

K j, i = K j, i ( y j, i , X j, i , Tj , P ) (2)

whereas, j, i denote, plate/component number respectively.

The optimal control problem can be described as follows: For a given pressure-swing batch distillation
configuration (N-total number of trays, P- working pressure), batch composition, the distillation task, and overall time
horizon (tf), determine the optimal reflux ratio, R = LD, to maximize the distillate, subject to any constraints (model
equations, bounds on the optimization variables).
The maximum distillate problem retained as:

Volume 5 Issue 2 |2024| 503 Fine Chemical Engineering

 min qN +1 ( t f ) (3)

dq
=f 0 (q ) + uf1 (q ), q ∈ X 1 (4)
dt

Hereby, the X1 denoted as reduced state space and defined as qN+1(tf)

X1 ={q =( X 1 , …, X N , U N , z ) : xi ∈ [0, 1]∀i } (5)

In the equation (4), q is composed of X1 ... XN, molar concentrations of the light component in the liquid phase, UN,
liquid hold-up in the reboiler, and z, purity deviation.

= ss
q0 q= , qN+2 ( t f ) 0 (6)

where is tf predetermined and qSS ∈ X 1 is composed of the steady state values of state variables. Vectors f0, f1 previously
defined by Stojkovic et al.25
As suggested by Modla,5 the boundary conditions to assure the infinite reflux condition for open mode rectifier:

0
(0) U=
U Nα = Nα U ch0 ; U iα =
(0) U i0 ; (7)

and to ensure the total reflux condition for the open-mode stripper are:

0
U N β (0) U=
= N β ; U iβ (0) 0; (8)

Whereas, the first/second column is denoted with the first/second letter of the Greek alphabet, respectively, and ch
denotes a respective charge content.

3. Numerical results and discussion

3.1 High-pressure driven process set-up
In this study, the investigation is expanded on a homogeneous azeotropic mixture, which thermodynamic properties
match very well with the thermodynamic model used (Figure 2). The working conditions and predefined parameters are
tabulated in Table 3, whereas concentrations are expressed from acetone, and LP/HP refers to low/high-pressure column,
respectively.

Table 3. Working conditions

Parameter P [atm] U N0 [l] U i0 [l] V [mol/h] tf [h] xN0 y*

LP: 1 11 LP: 0.75

Predefined/initial value 10 0.1 1.6 0.1
HP: 10 12 HP: 0.35

At the beginning of the dual process, the reboiler is filled with UN0 = 10 (l), and liquid hold-up on the trays is set
0
to U = 0.1(mol) for i = 0, ... N. For the boil-up, the constant value for the LP step is set to V1 = 11 (mol/h), however,
i

Fine Chemical Engineering 504 | Marija Stojkovic

at V2 = [11-13] (mol/h) for the HP step. The low-pressurized column works with the total number of trays N1 = 23,
whereas the high-pressurized column at even N2 = 13. As stated in the previous section, for the azeotropic kind of
mixture, the maximum achievable distillate concentration is limited by the position of the azeotropic point. Therefore,
the purities imposed for each step of the process: low/high-pressurized column to achieve, 75%/35%, respectively. Next,
the working conditions and predefined parameters are tabulated in Table 3, whereas LP/HP refers to low/high-pressure
columns, respectively.
Firstly, the operation is managed under standard atmospheric pressure, to distill acetone, i.e., 75% of left in the
accumulator. Consequently, the purified methanol (96.59%) is obtained in the reboiler and withdrawn. Furthermore,
the reboiler is heated by an additional auxiliary exchanger (Luyben17), before being pumped to the higher pressure
and introduced to the second column, this is necessary to increase boil-up. In the stripping step, the mixture near
the azeotropic mixture is collected in the condenser, to obtain purified methanol in the second tank. Moreover, the
interesting fact is that both, the purity and the recovery rate, tend to improve but at the cost of the total batch time. As a
consequence, it is a must to “repeat” a step to purify the next batch up to 95.50% and increase the recovery rate by an
additional 3% (Table 4). The total duration proposed for the best performance of the dual process is tf = 2.4 [h].
Despite the fact that the property of distilling under the pressure change, is exploited at maximum, the recovery
rate for the acetone is 72.83%, and, still almost 10% more than what was reported by Modla and Lang4 for the particular
configuration. In conclusion, the proposed operation is recommended to be employed as a two-step process for methanol
production and purification. Moreover, it is possible to have a second run for the HP step, where the acceptable purity
of 95% is achieved. The recovery rate improved by almost 4%, consequently, after the third run one can expect to reach
the maximum purity, but at the cost of the batch time increase, i.e., the total duration of 2.4 [h].

3.2 Numerical solution by BOCOP solver

A specific optimal control problem is formulated based on the assumptions of constant boil-up and product
specification for the respective step. Numerical solutions are brought by the BOCOP solver, as detailed by Bonnans
et al.29 This solver, in particular, employs the direct method with a choice of discretization schemes for dynamics time
discretization, which furthermore conversed the problem into finite-dimensional. In Table 4, the optimal costs and the
recovery rates of the respected component recovered in a two-step process are displayed.

Table 4. The optimal costs and the recovery rates

Product recovered N U0 (tf) Recovery rate Discr. scheme, nb. of points

Acetone 23 2.47865 72.83% Gauss 2,800

*
91.21%
Methanol 13 1.37305 Euler 2,800
94.47% **
* **
tf = 0.8 h, “single-period”, tf = 1.6 h, “double-period”

The optimal control policies for the azeotropic mixture of acetone-methanol are depicted in Figure 3, as the LP step
in the batch rectifier/HP step in the batch stripper, respectively.
For the first step, i.e., low-pressurized column (N1 = 23), as expected the initial and final arc match with maximum
distillate rate (infinite reflux condition, rectifier). It is to note that, the instantaneous pique that appeared after 0.47 [h],
will be disregarded. The optimal control trajectory is composed of 3 bang arcs of maximal distillate rate, but, even three
consecutive singular arcs with the tendency to reach approximately 27.30%/3.64%22.73% from the maximal distillate
rate. It is to note that, the latter mentioned, are taking over (more than any particular bang arc), approximately 23% of
the total time. In summary, the optimal control has an equal number of arcs of each type: 4 bang arcs, 3 arcs of zero
control, and, 3 singular arcs. Finally, there are 8 switchings between different types of optimal control.

Volume 5 Issue 2 |2024| 505 Fine Chemical Engineering

 12 20
(a) (b)
18
10
16
14
8
u(t) (mol/h)

u(t) (mol/h)
12
6 10
6
4
4
2
2
8
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
t, (h) t, (h)

Figure 3. The optimal distillate rate for u(t) for the: (a) LP step, (b) HP step

In the second step, i.e., high-pressurized process, as supposed the initial arc matches with zero distillate rate (total
reflux condition-stripper), but the final with maximal distillate rate. Further, it is observed that the optimal control
trajectory is composed of almost all bang-bang arcs, with two consecutive singular arcs preceding the final maximal
distillate arc. There are 13 commutations between different types of control, an interesting fact is that the singular arc (s)
are ‘positioned’ again in the penultimate period, whereas here it lasts approximately 6.25% of the total operation time
or more than 3 times shorter than in an LP step. In summary, the optimal control has an equal number of bang and zero
arcs, 6 and one singular arc.

64 138.5
(a) (b)
63 138
62 137.5
61 137
60 136.5
T (℃)

T (℃)

59 136
58 135.5
57 135
56 134.5
55 134
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
t, (h) t, (h)

Figure 4. The optimal temperature evolution for the: (a) LP step, (b) HP step

Moreover, in Figure 4, the output for the temperature evolutions along with the optimal solution is presented.
Above, the temperatures on all the stages of the column for the rectifier (LP step) are depicted. From here, the interval
of temperature increases along with the rectifier, going from the bottom to the top (batch to accumulator), following
the optimal trend from the minimal temperature of 55.25 °C to the maximal one at 63.31 °C, at the specified working
pressure of P1 = 1 atm. As previously stated, verify that the acetone is recovered in the product tank. Below, the
optimal temperature evolutions are presented for the HP step of the process, i.e., stripper, at the specified working
pressure of P2 = 10 atm, temperature rise, but in an inverse sense, from the top to the bottom, following the trend
within the temperature interval from the minimal temperature of 134.27 °C to the maximal temperature of 138.39 °C.
Hereby, previously written goes in favor of the fact that methanol is recovered in the inverted column product tank. All
previously stated are also following the findings reported by Luyben.16 As a consequence, it is verified, that the ultimate

Fine Chemical Engineering 506 | Marija Stojkovic

goal of the designed two-step process, i.e., consecutive production of acetone and methanol, is achieved.

3.2.1 The influence of liquid ratio

Focusing on the HP step, a short parametric study, whereas, the setup parameter of boil-up is varied, provided a
further study on the influence of the liquid division ratio on energy requirement. Data for the optimal liquid division
ratios obtained are tabulated in Table 5. Furthermore, after the validation of each pattern, the optimal reboiler duty
“pattern” was shown to be highly sensitive to the liquid division ration change, as if the latter varied by less than 9%, it
jumps for the next 19.81%.

Table 5. The liquid division ratio φ tabulated for different values of vapor boilup

L 3.182 3.311 3.533 3.543 3.652 3.756

φ 0.459 0.487 0.537 0.548 0.575 0.601

Additionally, the influence of vapor boil-up on the structure of the optimal control pattern is observed, as well. In
Figure 5, are plotted. Observations made, are as follows: the total number of cycles does not change for the decrease
of the vapor boil-up to 2 [mol/h], but further “gradual/stepwise” decrease results in the “equal decrease” of the total
number of cycles, since in the next step of the decrease by 4 [mol/h], the disappearance of the one cycle occurs; again
up to 6 [mol/h], the disappearance of even three cycles occurred, and up to 8 [mol/h], four cycles are disappeared.
The observations regarding the influence of the vapor boil-up on the optimal control pattern enable us to extend the
conclusions toward the prediction of the optimal heating strategy for each particular case. Finally, conclusions are drawn
about the total energy expenditure for the entire process, and the possibility of internal heat integration is thoroughly
examined, as discussed by Li et al.6
In Table 6, the investigation of the vapor boil-up is given, with different values of vapor boil-up, the achieved
distillate quantities, recovery rates, total number of cycles, and starting period duration.

Table 6. The investigation of the vapor boil-up

Vapor boilup, Distillate quantity, Total number of Staring period

Recovery rate
V [mol/h] [mol] U0 (tf) cycles, nc duration, [h]

7 3.2111 95.011% 4 0.4850

8 3.3017 93.427% 6 0.4195
10 3.5322 92.815% 8 0.3350
11 3,6423 92.185% 10 0.3000
12 3,7517 91.510% 12 0.2795
13 3,8556 90.853% 12 0.2595
14 3,9564 90.260% 14 0.2395
15 4,0539 89.660% 18 0.2195
16 4,1486 89.086% 18 0.2050
18 4,3285 87.971% 20 0.1825
19 4,4137 87.478% 22 0.1780
20 4,4960 86.961% 21 0.1600
* calculated per tf = 0.8 h, “single-period”

Volume 5 Issue 2 |2024| 507 Fine Chemical Engineering

 20
18
16
14
12

u(t) (mol/h)
10
8
6
4
2
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
t, (h)

Figure 5. The optimal control pattern for vapor boil-up varied

Compared to the previous cases, the operation should be initialized at higher values, however, the subsequent
almost instantaneous decreasing pique appeared again (Figure 6). It is perceived, that both consecutive “energy gaps”
are broadened”, this is followed by a very flattened decrease till the end of the first half of the total time. For the first
operational strategy proposed, the flattened area keeps its average value, but, almost double, compared to the second
one. This is followed by a more steep, since greater step, stepwise increase reaching the very last pique before switching
to the minimum of energy, followed by the final instantaneous increase of more than 3 times. On the other side, the
second proposed strategy, the “last” stepwise increase is instantaneously switched to the even greater pique, even on
average three times more compared to the first one. It is also important to mention that, the last pique of the increase
reaches more than 30% of the final estimated maximum, reboiler duty.

×104
3.5 6,000
(a) (b)
3 5,000

2.5
4,000
reboiler duty kJ/h

reboiler duty kJ/h

2
3,000
1.5
2,000
1

0.5 1,000

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
t, (h) t, (h)

Figure 6. Optimal reboiler duty vs time for vapor boil-up parameter increase as: (a) V = [12, 14, 16] [mol/h], (b) V = [18, 20] [mol/h]

It should be noted, however, that the “optimal reboiler duty” is a discontinuous time trajectory and, due to this,
could not be directly compared with the results obtained by Modla and Lang,9 since they used the continual heating
reboiler strategy over the total batch time. Despite this, one can perceive that for the same value of the fixed liquid
division ratio, the average energy required was at least 8.25% higher according to Modla and Lang9 compared to the
process proposed in this study.

Fine Chemical Engineering 508 | Marija Stojkovic

3.2.2 The influence of reboiler volume

20
18
16
14

u(t) (mol/h)
12
10
8
6
4
2
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
t, (h)

Figure 7. The optimal distillate rate u(t) for the HP step

Table 7. Working conditions

Parameter P [atm] U N0 [l] U i0 [mol] V [mol/h] tf [h] xN0 y*

0.8
Predefined/initial value HP:1 17.5271 0.1 20 0.1 LP: 0.75
1.6

Table 8. The optimal costs and recovery rates for acetone

Product recovered N U0 (tf) Recovery rate Discr. scheme, nb. of points

88.93%
Acetone (HP) 24 2.9588 Gauss 3,200
92.89%

Hereby, if the high-pressurized step, in particular, is “repeated” after its feed tank is filled with the additional
charge, a more complex pattern is for the optimal control structure. In the second step, i.e., high-pressurized process,
as supposed the initial arc matches with zero distillate rate (total reflux condition-stripper), but the final with maximal
distillate rate. Further, it is observed that the optimal control trajectory is composed of almost all bang-bang arcs, with
three consecutive singular arcs preceding the final maximal distillate arc. There are 30 commutations between different
types of control, since the three instantaneous piques achieving maximum distillate rate, are not taken into consideration
due to the extremely short period of duration. An interesting fact is that the singular arc (s) are ‘positioned’ again in the
penultimate period, whereas here it lasts approximately 7% of the total operation time or 8 times shorter than in an LP
step. The initial zero arc is of the longest duration, as it takes almost 40% of the operation time. In summary, the optimal
control has 15 bang arcs, 15 zero arcs, and three singular arcs (Figure 7). The working conditions, and, optimal costs
with recoveries, are tabulated in Table 7 and Table 8, respectively.

3.2.3 The influence of working pressure

Together with the influence of the reboiler volume, it is possible to examine the influence of high working pressure,
ranging between 6 and 10 [atm], i.e., pressure gap. The observations are tabulated in Table 9, for lower [6, 7 (atm)]
and intermediate pressure [8 (atm)], one case study is done, however, even two case studies are provided for the higher
pressures [9, 10 (atm)], notably for different charges: (1) 17.5271 [l], (2) 13.95 [l]. Consequently, the standard increase

Volume 5 Issue 2 |2024| 509 Fine Chemical Engineering

of 20% led to a productivity increase of 4.27%/4.89%, respectively, for the same imposed high pressures of 9/10 [atm],
respectively. All the more, keeping the first/second standard pressure constant, and increasing the higher pressure in
the second part of the process for not more than 10%, brings an improvement of 16%/17.37%, respectively, in the
yield. From the aspect of profit, one should consider the first option as a more profitable one. All optimal control policy
patterns are “near-BB” or “BB” (one short and pre-ultimate singular arc). It is to be noted that all the calculations are
done for the “single period” duration of the high-pressurized step, i.e., total batch time is set to a “single period” of time.

Table 9. The investigation of the influence of the pressure gap

Working pressure, P2 [atm] Distillate quantity, [mol] U0 (tf) Recovery rate Total number of cycles, nc Staring period duration, [h]
6 2.7890 91.08% 13 0.2020
7 2.6706 91.46% 12 0.2300
8 2.4165 92.27% 11 0.2780
2.4401 92.20% 11 0.2700
9
2.5041 89.87% 12-13 0.2640
2.9588 88.93% 15 0.1880
10
3.0391 87.74% 17 0.1860
* calculated per tf = 0.8 h, “single-period”

6,000 6,000
(a) (b)

5,000 5,000

4,000 4,000
reboiler duty kJ/h

reboiler duty kJ/h

3,000 3,000

2,000 2,000

1,000 1,000

0 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
t, (h) t, (h)

Figure 8. The optimal reboiler duty: (a) for LP step, (b) for HP step, for boil-up set at 11 [mol/h]

In this work, the number of strong assumptions on the fundamental dynamics of the column is reduced compared
to those made in the previous studies by Zhao et al.23 and Zhao et al.24, as it relies on the well-defined dependence of
the thermodynamic properties of the mixture from the temperature. However, to validate the results, ChemCad 6.1,
the dynamical simulator is employed, whereas the same applied thermodynamic model. Hereby, without any further
optimization, the interesting facts were retrieved for the change of heating energy during the respective process. In
Figure 8, the optimal reboiler duty vs time function for all stages was presented for the LP step, however, also depicts
the HP step: The extracted values for optimal reboiler duty vs time corresponding to the previously examined cases
of boil-up set at 11 [mol/h]. Compared to the energy required to heat the reboiler of the batch rectifier, it takes at least
14.88 times more to start up the batch stripper, from here the minimum total energy requirement is to be calculated for
the start-up of the proposed dual process: 60,400 kJ/h.

Fine Chemical Engineering 510 | Marija Stojkovic

4. Conclusions
The optimal control of a mixed configuration, comprising the batch rectifier and batch stripper, is determined in
terms of minimum distillate rate for a given amount of feed and quality. A rigorous model is built within the optimal
control framework in Bocop 2.2.1. Additional conditions on vapor-liquid equilibrium, necessary to accurately describe
high-pressure processes, are highlighted. Handled as thermodynamic constraints, some deeper insights into the
dependence of the optimal control structure on pressure effects are reported. The patterns of heating energy demand over
time during the process are developed: Compared to the previous research by Modla and Lang,9 in a double-column
batch stripper configuration, for the same fixed liquid division ratio, the average energy required was at least 8.25%
higher than that found for the process proposed in this study. Additionally, the authors did not discuss productivity,
whereas this study provides an overview of it. Moreover, investigating the influence of vapor boil-up on the optimal
control pattern showed that for an increase in vapor boil-up by 2.85 times, the total number of cycles increased by more
than five times, while the starting period duration decreased by 3 times. Furthermore, investigating the influence of
the pressure gap on the optimal control structure showed that for an increase in the pressure gap of 44.45%, the total
number of cycles increased by 23.53%, while the starting period duration decreased by 7.9%.
Given this, further research should focus on heat integration between different or the same units to achieve the
most acceptable outcomes in terms of production, energy, and profit. Moreover, delving deeper into optimal control
theory could extend ideas to newly established processes, such as multiple azeotropes treatment, unconventional
configurations/modes, and unified operations. Future research should also explore the introduction of entrainers to
investigate extractive column functioning and the recently presented entrainer-assisted pressure-swing distillation.
Moreover, none of the heat-integrated configurations is considered here, which should be in the future: thermally
integrated columns (DWC), internally integrated dividing-wall single columns (HIDIC), internally thermally coupled
dividing wall rectifier (TCDTBR), bottom flashing vapor recompressed batch distillation (BDBF), etc.

Conflict of interest
I declare there is no conflict of interest and the entire responsibility for the published results data is up to the author.

References
[1] Repke, J. U.; Klein, A. Homogeneous azeotropic pressure swing distillation: continuous and batch process. Comp.
Aid. Chem. Engng. 2005, 20, 721-726.
[2] Repke, J. U.; Klein, A.; Bogle, D.; Wozny, G. Pressure swing batch distillation for homogeneous azeotropic
separation. Chem. Eng. Des. Res. 2006, 85(4), 492-501.
[3] Modla, G.; Lang, P. New column configurations for pressure swing batch distillation I. Feasibility Studies. IFAC
Proc. Vol. 2007, 40(5), 105-107.
[4] Modla, G.; Lang, P. New column configurations for pressure swing batch distillation II. Rigorous simulation
calculations. IFAC Proc. Vol. 2007, 40(5), 361-366.
[5] Modla, G. Pressure swing batch distillation by double column systems in closed mode. Comp. Aid. Chem. Engng.
2010, 34(10), 1640-1654.
[6] Li, X.; Zhao, Y.; Qin, B.; Zhang, X.; Wang, Y.; Zhu, Z. Optimization of pressure-swing batch distillation with and
without heat integration for separating dichloromethane/methanol azeotrope based on minimum total annual cost.
Ind. Eng. Chem. Res. 2017, 56(14), 4104-4112.
[7] Kopacz, A.; Modla, G.; Lang, P. Product composition control of a new batch pressure swing rectifying system.
Hung. J. Ind. Chem. 2009, 37, 113-117.
[8] Kopacz, A.; Modla, G.; Lang, P. Operation and composition control of a new pressure swing batch distillation
system. Comp. Aid. Chem. Engng. 2009, 27, 1503-1508.
[9] Modla, G.; Lang, P. Separation of an acetone-methanol mixture by pressure-swing batch distillation in a double-
column system with and without thermal integration. Ind. Eng. Chem. Res. 2010, 49(8), 3785-3793.
[10] Modla, G. Separation of a chloroform-acetone-toluene mixture by pressure-swing batch distillation in different

Volume 5 Issue 2 |2024| 511 Fine Chemical Engineering

 column configurations. Ind. Eng. Chem. Res. 2011, 50(13), 8204-8215.
[11] Modla, G.; Lang, P. Comparison of extractive and pressure-swing batch distillation for acetone-methanol
separation. Comp. Aid. Chem. Engng. 2011, 29, 382-386.
[12] Modla, G. Reactive pressure swing batch distillation by a new double column system. Comp. Aid. Chem. Engng.
2011, 35(11), 2401-2410.
[13] Li, X.; Yang, X.; Wang, S.; Yang, J.; Wang, L.; Zhu, Z.; Cui, P.; Wang, Y.; Gao, J. Separation of ternary mixture
with double azeotropic system by a pressure-swing batch distillation integrated with quasi-continuous process.
Process Saf. Environ. Prot. 2019, 128, 85-94.
[14] Wang, Y.; Yang, X.; Zhao, J.; Liu, X.; Yao, D.; Cui, P.; Wang, L.; Zhu, Z.; Li, X.; Xu, D. Design and comprehensive
analysis of a novel pressure-swing batch distillation process for the separation of a binary azeotrope with various
boiling behaviours. Sep. Purif. Technol. 2020, 251, 117329.
[15] Luyben, W. L. Optimum vacuum distillation pressure. Chem. Eng. Process.: Process Intensif. 2021, 171, 108630.
[16] Luyben, W. L. Pressure-swing distillation for minimum- and maximum-boiling homogeneous azeotropes. Ind. Eng.
Chem. Res. 2012, 51, 10881-10886.
[17] Luyben, W. L. Importance of pressure selection in pressure-swing distillation. Comp. Aid. Chem. Engng. 2021,
149, 107279.
[18] Kale, P.; Pujari, S.; Gujar, J. G.; Sontakke, R.; Haddadi, E. I.; Sonawane, S. S. Batch distillation for separating the
acetone and n-heptane binary azeotrope mixture: Optimization and simulation. J. Indian Chem. Soc. 2023, 100(1),
100795.
[19] Ding, Z.; Zhang, H.; Li, H.; Chen, J.; Lu, P.; Hua, C. Improved Harmony Search Algorithm for Enhancing
Efficiency and Quality in Optimization of the Distillation Process. ACS Omega. 2023, 8(31), 28487-28498.
[20] Hegely, L.; Karaman, Ö. F.; Szucs, M. T.; Lang, P. Surrogate model-based optimization of a batch distillation
process. Chem. Eng. Res. Des. 2023, 192, 456-467.
[21] Qin, Y.; Zhuang, Y.; Wang, C.; Zhang, L.; Liu, L.; Du, J. Multi-objective optimization and comparison of the
entrainer-assisted pressure-swing distillation and extractive distillation separation sequences for separating a
pressure-insensitive binary azeotrope. Comp. Chem. Engng. 2022, 165, 107959.
[22] Miao, G.; Zhuo, K.; Li, G.; Xiao, J. An advanced optimization strategy for enhancing the performance of a hybrid
pressure-swing distillation process in effective binary-azeotrope separation. Sep. Purif. Technol. 2022, 282, 120130.
[23] Zhao, J.; Shen, Y.; Li, C.; Zhao, F.; Li, X.; Zhu, Z.; Wang, Y.; Cui, P.; Gao, J. Sequential two-column batch
distillation processes for separation of a ternary mixture containing three binary minimum boiling point
homoazeotropes. Sep. Purif. Technol. 2021, 270, 118826.
[24] Zhao, J.; Zhu, Z.; Ma, Z.; Zhao, F.; Yang, X.; Wang, Y.; Cui, P.; Li, X.; Gao, J. Double-column batch stripper
process based on heterogeneous property and control strategy for the efficient separation of a ternary mixture
containing two minimum boiling azeotropes. Process Saf. Environ. Prot. 2021, 148, 1123-1132.
[25] Stojkovic, M.; Gerbaud, V.; Shcherbakova, N. Cyclic operation as optimal control reflux policy of binary mixture
batch distillation. Comp. Aid. Chem. Engng. 2018, 108, 98-111.
[26] Doherty, M. F.; Malone, M. F. Conceptual Design of Distillation Systems; New York: McGraw-Hill, 2001; pp 568.
[27] Gmehling, J.; Onken, U. Vapor-Liquid Equilibrium Data Compilation: DECHEMA Data Series; DECHEMA,
Frankfurt, 1977.
[28] Smith, J. M.; Van Ness, H. C. Introduction to Chemical Engineering Thermodynamics; Singapore: McGraw Hill,
1996; pp 769.
[29] Bonnans, F.; Giorgi, D.; Grélard, V.; Maindrault, S.; Martinon, P. User Guide BOCOP. https://www.bocop.org
(accessed January 06, 2023).

Fine Chemical Engineering 512 | Marija Stojkovic

View publication stats